Document 10677181

advertisement
Applied Mathematics E-Notes, 3(2003), 16-37 c
Available free at mirror sites of http://www.math.nthu.edu.tw/∼amen/
ISSN 1607-2510
A Survey On Homoclinic And Heteroclinic Orbits
∗†
Bei-ye Feng and Rui Hu‡
Received 2 January 2002
Abstract
The study of homoclinic and heteroclinic orbits has a long history. This paper
surveys some of the recent advances in this topic. We present some of the main
results obtained in recent years, and in addition we indicate some possible research
directions and some problems for further studies.
1
Introduction
Homoclinic and heteroclinic orbits arise in the study of bifurcation and chaos phenomena (see e.g. [1—7], [10], [48, 49] and [55]) as well as their applications in mechanics,
biomathematics and chemistry (see e.g. [29], [108—111]). Many works related to these
topics have been done in recent years. In this paper, we intend to survey some of the
results, methods and problems that have recently been reported.
Consider the dynamical system
Ẋ = f (X), t ∈ R,
(1)
where X ∈ Rn , f ∈ C r and r ≥ 1. Recall that a point X0 is a singular or equilibrium
point of (1) if f (X0 ) = 0. A singular point X0 is a hyperbolic point if all of the
eigenvalues of Df (X0 ) have nonzero real parts and is an elementary saddle point if
det Df (X) |X=X0 < 0. Let p, q be hyperbolic points of the above system (1). A
solution path sp : X = X(t) of (1) is called a homoclinic orbit connecting with p if
X(t) → p as t → ±∞ (See Figure 1),
Figure 1
∗ Mathematics
Subject Classifications: 34C05, 34C23, 34C37.
research is supported by the NNSF of China (No.10171099)
‡ Institute of Applied Mathematics, Academy of Mathematics and Systems Sciences, Chinese
Academy of Sciences, Beijing 100080, P. R. China
† This
16
B. Y. Feng and R. Hu
17
and a solution path spq : X = X(t) of (1) is called a heteroclinic orbit connecting with
p and q if X(t) → p as t → −∞ and X(t) → q as t → +∞ (See Figure 2).
Figure 2
n
the
Some heteroclinic orbits may form a cycle. We will denote by SO
1 ,O2 ,···,On
heteroclinic cycle passing through the singular points O1 , O2 , · · · , On (See Figure 3).
Figure 3
There are several natural aspects related to these orbits which are worthy of discussions:
I. The existence of homoclinic and heteroclinic cycles.
II. The stability of a homoclinic or a heteroclinic cycle.
III. The mutual position between the stable and the unstable manifold of a saddle
point.
IV. The bifurcation of a homoclinic or heteroclinic orbit.
2
The Existence of Homoclinic and Heteroclinic Cycles
The research of this problem provides the foundation for the research of the others.
There are several well known methods dealing with the existence problems.
The first method is to study a planar Hamiltonian system, because it is basically
an analytic geometry problem and it is comparatively easy to determine the existence
of the homoclinic cycles or heteroclinic cycles of Hamiltonian systems.
The second method is the so-called structured approach. For example, consider the
plane system
ẋ = P (x, y),
(2)
ẏ = Q(x, y).
Select a function F (x, y) such that F (x, y) = 0 for all (x, y) which belongs to a closed
curve L passing through a saddle point (i.e. a singular point (x0 , y0 ) of (2) satisfying
18
Homoclinic and Heteroclinic Orbits
det ∂(P (x0 , y0 ), Q(x0 , y0 ))/∂(x, y) < 0) and
dF
|(2) = Fx P + Fy Q = F B
dt
where B is a regular (or analytic) function in the neighborhood of L. Then we can
conclude that there exists a saddle separatrix cycle L. This method can work for nonHamiltonian system. Furthermore, by constructing a rotational vector field [7], we can
also get a saddle separatrix cycle of a system. For example, let L : F (x, y) = 0 be the
separatrix cycle of system (2), then we can conclude that L is the separatrix cycle of
the system
ẋ = P (x, y) + F P1 (x, y),
(3)
ẏ = Q(x, y) + F Q1 (x, y).
Using this method, we can get a complicated vector field in which there exists a separatrix cycle through a simple vector field in which there exists a separatrix cycle. There
are many examples given in [7], [8], [9], [13], [15], [16], [20], [21], [22], [37], [38], [43],
[50], [58].
The third method is to study some special separatrix cycle such as the separatrix
cycle formed by the conic curve of a quadratic system. This method can be found in
[7], [59] and [60]. It is also important to study the existence of separatrix cycles of
systems such as the Lorenz system. Many results about the existence of the separatrix
cycles, whose solution path cannot be given explicitly, are obtained by perturbation
methods (see for examples [37] and [38]).
3
The Stability of a Homoclinic or a Heteroclinic
Cycle
Let us denote the system (2) by I. This is convenient since we will introduce a related
system later which will be indicated by I(ε). dependence on ε. Assume that there is a
homoclinic cycle L passing through the saddle point O(0, 0) in system (2).
In the investigation of the stability of a homoclinic cycle, Dulac in 1923 studied the
analytic system by using the so-called semi-regular function and obtained the following
result.
THEOREM 1. In system (2), assume that P and Q are analytic. Then L is stable
(unstable) if σ0 = (Px + Qy )|(x,y)=O(0,0) < 0 (respectively > 0).
The proof is rather complicated. In 1958, by using the success function, Andronov
et al. assumed the system is C 1 and proved the theorem mentioned above in [2]. Under
the assumption that P and Q are C 1 , Melnikov [11] gave a proof of this theorem in
τ
1966. Chow proved this result by another method in [15]. The sign of 0 div(I)|(x,y)∈γ dt
(where γ is a path near L, and τ is the time in which γ finishes a return or Poincare
map) were used to determine the stability of L in [11] and [15], and this term was
obtained when Andronov et al. [2] computed the success function. Therefore, the work
of [2] is the base of the others.
When σ0 = 0, the first critical case appears. Andronov et al. [3] first considered
this problem. From the proof of Theorem 1 in [11], we may infer the reason why the
B. Y. Feng and R. Hu
19
stability of L can be determined by the information of the saddle point itself when
σ0 = 0. Andronov et al. [3] once tried to expand the saddle point by the Taylor
expansion, and use the first non-zero term of the expansion to determine the stability,
but finally, they gave a counter-example instead. Today, it is well known that for the
first critical case it is impossible to determine the stability of the homoclinic cycle only
by using the information of the saddle point itself.
The following result is given in [3].
LEMMA 1. σ1 =
(respectively σ0 = 0).
+∞
(Px
−∞
+ Qy )|(x,y)∈L dt is convergent (+∞ or −∞) if σ0 = 0
This Lemma enables us to state the criterion for determining the separatrix cycle
for the critical case in simpler form. In analogy with the fact that the stability of a
limit cycle Γ is determined by the integral Γ div(I)dt, it is natural to guess that the
+∞
integral −∞ div(I)|(x,y)∈L dt can be used to determine the stability of the homoclinic
cycle L. In 1985, using the C 1 -Hartman theorem [77], Feng and Qian in [4] proved the
following result.
THEOREM 2. If σ0 = 0 and σ1 =
stable (respectively unstable).
+∞
(Px
−∞
+ Qy )|(x,y)∈L dt < 0 (> 0), then L is
We remark that Luo et al. simplified the proof of this theorem in [5] (see also [11],
[79]). The work of [4] has the following significance: 1. The stability of a homoclinic
cycle in the critical case must be determined by all the information of the homoclinic
cycle instead of by the information of the saddle point itself. 2. A new method is
presented (dealing with success function in the regular domain and singular domain
respectively) which is later used in [17], [22] and [38].
When σ0 = σ1 = 0, the second critical case appears. Han and Zhu [73] pointed
out that the first order saddle quantity can determine the stability of L for this case.
In 2001, by using the property that the successor function is intrinsic and it has no
relation with the curvilinear coordinate system, Hu and Feng [72] obtained the limiting property of the successor function by constructing a series of special curvilinear
coordinate systems, and consequently, they obtain the criterion below.
THEOREM 3. Assume that in system (2), σ0 = σ1 = 0. Let
+∞
σ2 = −
t
e
−∞
0
(Px +Qy )|L dτ
· V |L dt
(4)
where
V
=
1
[(3Px Qx + Px Py + Qx Qy + 3Py Qy )(Q2 − P 2 )
(P 2 + Q2 )2
+2P Q(Py2 − Q2x + 2Px2 − 2Q2y )]
1
+ 2
[−Q(2Pxx + Qxy + Pyy ) + P (Pxy + Qxx + 2Qyy )].
P + Q2
If σ2 < 0, then L is stable from inside (so that every solution curves near and inside L
tend to L as t → +∞). If σ2 > 0, then L is unstable from inside.
20
Homoclinic and Heteroclinic Orbits
The integral on the right hand side of (4) equals infinity when the first order saddle
quantity does not equal 0, so the result in [72] also pointed out that the first order saddle
quantity can determine the stability of the homoclinic cycle for the second critical case.
Luo and Zhu [22] gave some examples to point out that it is useful to study the higher
critical cases.
As for the stability of heteroclinic cycles, Cerkas [14] in 1968 obtained the following
theorem.
THEOREM 4. Assume that in system (2), S (n) is the heteroclinic cycle passing
through the elementary saddle points O1 , O2 , · · · , On , and
−
λ+
i > 0 > λi
are the eigenvalues of the system at the saddle point Oi . Let
λi = −
λ−
i
, λ = λ1 λ2 · · · λn .
λ+
i
Then, when λ > 1 (< 1), S (n) is stable (respectively unstable).
When n = 1, the heteroclinic cycle degenerates into a homoclinic cycle S (1) , and
the criterion here is that when λ = −λ− /λ+ > 1 (< 1), S (1) is stable (respectively
unstable). In view of σ0 = λ+ + λ− , it is clear that
σ0 < 0 (> 0) ⇐⇒ λ > 1 (< 1).
This fact shows that Theorem 2 is a special case of Theorem 4.
The results of [14] not only gave the criterion for determining the stability of a heteroclinic cycle, but also change the thoughts of some mathematicians. σ0 is important
for the study of a homoclinic cycle, so, some people want to determine the stability
(1)
(2)
(n)
(i)
of S (n) by σ0 , σ0 · · · , σ0 . From Theorem 4, we learn that instead of σ0 , λ+
i and
−
λi are the essential quantities, and the reason that σ0 can be used to determine the
stability of S (1) is that σ0 happens to be equivalent to λ when n = 1.
When λ = 1, the critical case for a heteroclinic cycle appears. Some mathematicians
once suspected that the integral ( s12 + s23 + · · · + sn1 )(div(I))dt (where sij is the
heteroclinic path from the saddle point Oi to the saddle point Oj ) can be used to
determine the stability of S (n) . But in fact, this conjecture can be right only if λ1 =
λ2 = · · · = λn = 1, so it is a special case. Before the right answer has been given, many
people may be misled by this conjecture.
By means of the methods used in [12], [14] and [15], Feng [17] obtained the following
result.
THEOREM 5. In system (2), let Mi1 , Mi2 be the points on the stable manifold s+
i
and unstable manifold s−
i , and the arc length Oi Mi1 = Oi Mi2 = ci ρ (where ρ > 0 is
small). Then, S n is stable (unstable) if λ > 1 (respectively < 1) or λ = 1 and Λn1 < 0
(> 0), where
Λn1 = lim (Jn1 + λn Jn−1,n + · · · + λ2 · · · λn J12 ),
ρ→0
Jij =
div(I)dt, j = i + 1,
Mi2 Mj1
and when i = n, j = 1,
B. Y. Feng and R. Hu
21
and ci are determined by the equations
1
(λ+
c2−λ
1 ) cos ϕ1
1
2−λ2
c2
(λ+
2 ) cos ϕ2
....
n−λn
+
cn
(λn ) cos ϕn
=
=
=
=
c2 (−λ−
2 ) cos ϕ2
c3 (−λ−
3 ) cos ϕ3
...
c1 (−λ−
1 ) cos ϕ1
here ϕi = |90◦ − θi |, and θi is the angle subtended by si−1,i and si,i+1 .
When λ = 0, corresponding to n = 1 and n > 1, Theorem 5 is equivalent to
Theorems 2 and 4, when λ = 1 and n = 1, Theorem 5 yields Theorem 3, and when
λ1 = λ2 = · · · = λn = 1, Theorem 5 extends Theorem 2.1 in [25], i.e. the conjecture
above. So the results about the stability of a homoclinic cycle and heteroclinic cycle
in [1]-[3], [12], [14] and [23]-[25] are special cases of the results in [17]. Theorem 5 not
−
only contains the quantities λ+
i , λi , λi , sij div(I)dt, but also indicates that the brevity
of the criterion depends on the special proportion between the positions of the arcs
which are not tangent with any solution paths (i.e. arcs without contact). (The proof
in [14] shows that if the proportion between the positions of the arcs without contact
is not appropriate, the complicated expression would appear in the criterion so that
the criterion cannot be applied.) When it comes to the convergence of the integral
in the theorem, new situations about the heteroclinic cycle arise. Lemma 1 indicates
that L div(I)dt is convergent if σ = 0 or λ = 1. Feng [17] proved that this property
is right for n = 2. In other words, the integrals s12 div(I)dt and s21 div(I)dt are
convergent, but the convergence here is weaker than that when n = 1, and it is based
on the proportion between the positions of the arcs without contact in Theorem 5 and
is the convergence under Cauchy’s principal value. When n ≥ 3, we can only guarantee
the convergence of some special linear combination of them instead of the convergence
under Cauchy’s principal value. Feng [41] gave a brief survey about the study of this
problem.
All the theorems above assume that the parametric equations of the homoclinic
or heteroclinic cycles are all known. But in many cases, it is impossible to give the
parametric equations of the homoclinic cycle or heteroclinic cycle. In [8], [9] and [58],
the authors studied some special systems. When Zhang [58] proved the stability of
a heteroclinic cycle, this problem is changed into the uniqueness of the heteroclinic
cycle, i.e. to prove that there does not exist a limit cycle in the neighborhood of
the heteroclinic cycle, and then the stability of the singular point can determine the
stability of the heteroclinic cycle.
Another problem is the stability of an infinity separatrix cycle. An infinity separatrix cycle is a solution path s∞ both of its ends tend to infinity as t → ∞ (See Figure
4).
Figure 4
22
Homoclinic and Heteroclinic Orbits
The stability of s∞ is taken to mean that the solution paths near s∞ approach it
when t → ∞. In [26], [27] and [28], a quadratic system is considered, and its infinity
separatrix cycle whose interior has a focus point is studied. Feng [13] studied the infinity
separatrix cycle whose interior has a center. In [39], Feng also studied the Melnikov
functions of an infinity separatrix cycle. And finally, in 1995, Feng [70] obtained the
criterion for determining the stability of a separatrix tending to infinity as follows.
THEOREM 6. Consider the system
ẋ = P0 (x, y),
ẏ = yQ0 (x, y),
(5)
where y = 0 is the separatrix tending to s∞ . (Remark: The system (2) can be turned
into this form by means of a transformation) Assume that (i) a return map can be defined in the neighborhood of the upside of y = 0, (ii) in system (5), P0 = (P0 −P∞ )+P∞ ,
Q0 = (Q0 −Q∞ )+Q∞ , where P∞ , Q∞ has the properties (H1 ) yQ∞ /P∞ is the homogeneous function of x, y, (H2 ) (P0 −P∞ )/P∞ → 0 and (Q0 −Q∞ )/Q∞ → 0 as x2 +y 2 → 0,
+∞
(iii) the integral I1 = −∞ Q0 (x, 0)/P0 (x, 0)dx, interpreted as Cauchy’s principal
+∞
∞ +cP∞
value, is convergent, (iv) the integral I2 = −∞ (1−c)xQ
x(xQ∞ −P∞ ) (x, 1)dx, interpreted
as Cauchy’s principal value, is convergent, where c = Q∞ (1, 0)/(Q∞ (1, 0) − P∞ (1, 0)).
Then the separatrix tending to s∞ is stable (unstable) upside if σ∞ = I1 +I2 /(1−c) < 0
(respectively > 0).
A concept which is important in discussing the stability of a space separatrix cycle
is given by Feng [71] in 1996.
DEFINITION 1. Let A be a 3 × 3 matrix. If A has three real eigenvalues, which
can be denoted as λ− < 0 < λ+ and λ∗ such that the tangent vectors b+ and b− of S (1)
at the saddle O are the characteristic vectors corresponding to λ− and λ+ respectively,
whereas λ∗ is the remaining third eigenvalue. If A has a pair of complex conjugate
eigenvalues and one real eigenvalue, then the unique real eigenvalue is denoted as λ∗ .
+
∗
In a similar way we can define the concept of λ−
i , λi and λi .
A stability criterion is given below for a homoclinic cycle (and a corresponding
criterion for a heteroclinic cycle is similar to it).
THEOREM 7. Consider the system
Ẋ = F (X), X ∈ R3 , F ∈ C 2 ,
(6)
which admits a homoclinic cycle S (1) and the saddle point is the origin O. Assume
that (i) the eigenvalues of A = DX F (O) are all real, (ii) for some ε0 > 0, S (1) has a
positive direction invariant ε0 -part neighborhood, (iii) λ∗ < 0, λ = −λ− /λ+ > 1 (< 1),
then, S (1) is positively asymptotic stable (respectively unstable).
In the above result ,we say that U is a positive invariant 0 -neighborhood of a
solution ϕ(X0 , t) of (6) passing through X0 if {ϕ(X0 , t) | X0 ∈ U } ⊂ U for any t and
d(U , S (1) ) < 0 for any X0 ∈ U .
We remark that it is an interesting problem to determine the stability of S (1) when
λ = 0.
B. Y. Feng and R. Hu
23
By extending Theorem 7 to high dimensional case, we give an application in biology.
The results we give extend those obtained by May and Leonard in [109] and by Hofbauer
and Sigmund in [110].
All the problems above have relations with the normalized form of the system in the
neighborhood of the saddle point (see [85]). For example, in [12], [14] and [23], the C 1 Hartman Theorem, invariant manifold orthogonalizing theorem and Joyal normalized
form theorem are employed for obtaining their results.
Normalized form was studied for a long time and it has relations with the smoothness of the system, the property of the eigenvalue and the smoothness of the transformation. For example, a system which cannot be linearized by means of an analytic
transformation may be linearized by means of a C 1 transformation. In [12] and [14],
the systems are assumed to be C 2 , and in [23], the systems C 3 . Although there were
many profound results about the normalized form, there are some unsettled problems.
For instance, the famous Poincare-Sternber Theorem says that if the eigenvalues of
a planar system are nonresonant, then this system can be C ∞ linearized. But there
also exist some systems which do not satisfy the nonresonant conditions, yet can be
linearized by means of analytic transformations. For example, the system
ẋ = −y + xy
ẏ = x + y 2
can be linearized by means of the transformation
u =
v
=
x
,
1−x
y
.
1−x
So, the conditions of linearization are worthy of studying. For related studies, the
reader may consult [4], [5], [12], [14], [16],[23], [33], [45] ,[60]-[63] and [94].
4
The Mutual Position Between the Stable and the
Unstable Manifold of a Saddle Point
In 1963, Melnikov [11] first obtained an important result which provides an analytic
criterion for determining the mutual position of the separatrixes.
This method has been applied widely and is called the Melnikov method. The
determining function in this method is called Melnikov function. A brief sketch of this
method is as follows.
Consider the system
ẋ = P (x, y) + εP1 (x, y, )
(7)
ẏ = Q(x, y) + εQ1 (x, y, ).
Let us denote the system (7) by I(ε) to indicate its dependence on ε. There exists
separatrix cycle L0 in system I(0). Furthermore, when the system is perturbed, L0
“breaks” and becomes the stable manifold Lεs and the unstable manifold Lεu (See
Figure 5). The problem is to determine the mutual position of Lεs and Lεu . Take the
24
Homoclinic and Heteroclinic Orbits
points Mt+ on Lεs and Mt− on Lεu , then, we get a vector Mt− Mt+ depending on the
time t and the parameter ε. It is clear that the direction of this vector can determine
the mutual position of Lεs and Lεu . The information about the system I(0) is known,
but it does not contain the information about Mt− Mt+ , so it is almost impossible to
study the Mt− Mt+ itself to get the mutual position of Lεs and Lεu . Melnikov’s method
is to take the exterior normal vector n(t) corresponding to the time t on the separatrix
cycle of the unperturbed system I(0), then by comparing the directions of Mt− Mt+ and
n(t), we can determine the mutual position of Lεs and Lεu , i.e. when Mt− Mt+ and n(t)
have the same directions, Lεs is on the outside of Lεu , otherwise, Lεs is on the inside
of Lεu . Therefore, it is natural to consider the inner product of Mt− Mt+ and n(t), i.e.
to consider the following function due to Melnikov
∆ε (t) = −n(t) · Mt− Mt+ .
The function ∆ cannot be computed. For any vectors α = (α1 , α2 ) and β = (β1 , β2 ) in
2
, define α ∧ β = (−α1 , α2 ) · (β1 , β2 ) . The next step is to prove the expansion
∆ε (t) = ∆1 (t)ε + ∆2 (t)ε2 + · · · ,
and to prove that the differential equation below is satisfied by ∆1 (t)
d∆1 (t)
dt
(8)
= σ(t)∆1 − D(t),
where
σ(t) = div(P, Q)|x=ϕ(t),y=ψ(t) ,
and
D(t) =
P
Q
∧
P1
Q1
x=ϕ(t),y=ψ(t),ε=0,
here x = ϕ(t) and y = ψ(t) are the parametric equations of L0 .
Figure 5
Thus, when ε is sufficiently small, we can get the sign of ∆ε through the sign of
∆1 , where ∆1 can be computed. In view of (8), Melnikov asserts that
+∞
∆1 =
−
[D(t)e
−∞
t
0
σ(ξ)dξ
]dt.
B. Y. Feng and R. Hu
25
This is the so-called Melnikov criterion. However, Feng and Qian in [12] found that
(8) can only yield
−
+
∆1 (0) = ∆+
1 (T )e
T+
0
σ(t)dt
−
−
− ∆−
1 (T )e
T−
0
σ(t)dt
T+
+
[D(t)e
−
t
0
σ(ξ)dξ
]dt. (9)
T−
As a consequence, Melnikov function can be used as the first order approximation only
when the first two terms on the right hand side (called the remainder terms) in (9)
tend to zero as T + → +∞ and T − → −∞. In [12], it is proved that when the saddle
point is elementary (when the two eigenvalues are both nonzero), the remainder terms
tend to zero as T + → +∞ and T − → −∞. We believe that Melnikov assumed that
the saddle point of the system was elementary, and thus the terms tend to zero. But
for the higher order singular point, a counter-example is given in [12] to illustrate that
the remainder terms do not tend to zero. This prompts us to consider conditions which
are sufficient for the remainder terms to vanish. For instance, in [13], such a condition
is given for the system which has the homogeneous main part, i.e.
ẋ = P + εP1 ,
ẏ = Q + εQ1 ,
P = R1 + f1 ,
Q = R2 + f2 ,
(10)
where P = P (x, y), Q = Q(x, y), P1 = P1 (x, y, ε), Q1 = Q1 (x, y, ε), R1 and R2 are the
m-th order homogeneous polynomials, P1 and Q1 have order greater than or equal to
m, f1 and f2 have order greater than or equal to m + 1. Another general result is given
in [39].
THEOREM 8. Assume that (a) the system (10) when ε = 0 is a Hamiltonian
system, or (b) the saddle point of (10) when ε = 0 is hyperbolic, i.e. det A < 0, where
A is the linearized matrix of the system at the point O, or (c) m + 1 − |π ∗ | > 0, where
π± =
R1x + R2y
|x=1,y=k± ,
R1
and k+ or k− is the rate of slope of the stable manifold (k+ ) or the unstable manifold
(k− ) at the point O. Then the remainder terms vanish.
Under the conditions of the Theorem above, we can use the Melnikov criterion as
the first order approximation of the Melnikov function. At the same time, we can
determine the mutual position of separatrix cycle in the perturbed system with ∆1 (0).
Feng and Qian [12] presented the following theorem.
THEOREM 9. Assume that ε is sufficiently small, then if ε · ∆1 (0) < 0 (> 0), L+
ε
is on the outside (respectively inside) of L−
ε (see Figure 6).
We remark that for systems which have nonhomogeneous main parts, the vanishing
conditions for the remainder terms have not been studied. We remark further that for
systems which have homogeneous main parts, the result of [13] does not include the case
m = 1, i.e., the hyperbolic situation. Another interesting problem is the computation
of the remainder terms when they do not vanish.
26
Homoclinic and Heteroclinic Orbits
Figure 6
When ∆1 (0) = 0, the first order approximation cannot be used to determine the
sign of ∆ε (0), so we must consider the so-called higher order Melnikov function. Yuan
[40] and Sun [50] studied this problem, and gave the expressions of the second order
Melnikov function. But their expressions are different. Yuan [40] obtained the criterion through computing the right-hand member of the system, and Sun [50] obtained
the criterion by the solution of a linear variational equation. No one has offered a
comparison of these two methods yet.
All the problems above involve small perturbations, i.e. it is assumed that there
exists a separatrix cycle in the unperturbed system. We also want to find out the mutual
position of the two separatrixes in the perturbed system. For the “large region” case,
there are few results. Even the formulation of the problem is unclear. We have tried
to formulate the problem as follows.
Consider the system (2), assume that O(0, 0) is the saddle point of the system. (a)
Under what conditions will the stable manifold Ls and unstable manifold Lu of O all
intersect with some arc without contact of the system? (b) If Ls and Lu all intersect
with some arc without contact, what are the conditions for determining their mutual
positions.
Dumortier [55] and Zhou [56] have studied this problem. But it is difficult to give
a “good” condition such that when the system (2) turns into the system I(ε), this
condition reduces to the Melnikov condition. But there is a question here. In the
system I(ε), the information of the separatrix cycle L0 of I(0) is known, but in system
(2), we cannot take Ls and Lu as the known quantities. So, when the system (2)
becomes I(ε), we know more information than before. However, we cannot express
this process. In other words, this problem has not been settled, and many important
problems depend on the determination of the mutual position of the separatrixes.
5
The Bifurcation of a Homoclinic or a Heteroclinic
Orbit
There are several problems related to the study of bifurcation:
(1) to determine the types of the bifurcation,
B. Y. Feng and R. Hu
27
(2) to determine the number of limit cycles due to bifurcation,
(3) to determine the stability of bifurcation,
(4) to give the conditions for determining the problems above.
In [3]-[5],[48], [49] and [63], the authors have studied the first aspect thoroughly and
considered almost all the possible cases. Recently, the authors of [34], [44], [46], [54]
and [55] also did many works. Yet, they all restricted their studies in the possibility of
the bifurcation, and did not consider the condition of the bifurcation. But determining
the condition of bifurcation may never be settled completely. The author of [49] has
pointed out that many bifurcation sets are not algebraic, i.e. it is impossible to give the
algebraic determinant. So we see there are many bifurcation conditions which cannot
be given. Li et al. [80] also did some related works.
As for the second problem, in 1958, Andronov [3] proved that if σ0 = 0, the limit
cycle which bifurcates from the separatrix cycle has the same type of stability with the
original separatrix cycle, so this limit cycle must be unique. In [3], a counterexample
is also given to show that when σ0 = 0, the separatrix cycle can bifurcate into two
limit cycles. So with this result in mind, many people had thought that there was no
uniqueness when σ0 = 0. To their surprise, Luo and Zhu [22] assumed that σ0 (ε) ≡ 0
and proved that when σ0 = 0 and σ1 = L div(I)dt = 0, the limit cycle has the
same stability with the separatrix cycle from which it bifurcates (and hence is unique).
In fact, the integrals Γ divI(ε)dt and L div(I)dt determine the stability of the limit
cycle and the separatrix cycle respectively. While I(ε) → I and Γ → L as ε → 0, so
divI(ε)dt → L div(I)dt as ε → 0. In the same manner, Hu and Feng [72] proved
Γ
that under appropriate conditions, when σ0 = σ1 = 0 and σ2 = 0, there exist at most
2 limit cycles bifurcating from L in the neighborhood of L. The newest results about
the uniqueness of bifurcation can be seen in [91].
For the number of the bifurcation limit cycles, there are some results about Hamiltonian systems. In 1951, Leontovich [30] gave the result below without proof. R.
Roussarie [53] gave the same result with proof. Consider the system
ẋ = Hy + εf (x, y),
ẏ = −Hx + εg(x, y).
(11)
Let I(ε) denote the system (11). Assume that Lh is the contour line H(x, y) = h of
system I(0). L0 is the separatrix cycle, H > 0 denotes the interior of L0 and
I(h) =
(
H≥h
∂f
∂g
+
)dxdy.
∂x ∂y
THEOREM 10. When h is sufficiently small,
I(h) = c1 + c2 h ln h + c3 h + c4 h2 ln h + · · · ,
where
∂f
∂
∂g
+
)|x=y=0 , c3 =
I(h)|h=0 .
∂x ∂y
∂h
If I(h) = 0 and c1 = c2 = · · · = cn = 0, cn+1 = 0, then there exists a perturbed system
of I(ε), and there exist n limit cycles near L0 in this system. Furthermore, there exist
at most n limit cycles in the perturbation of I(ε).
c1 = I(0), c2 = (
28
Homoclinic and Heteroclinic Orbits
W. G. Li proved the convergence of the above expansion in [94].
Based on the theorem above, the authors of [20] and [21] proved that there exist at
most 3 limit cycles bifurcating from the quadratic system. Rousseau [64] first showed
the existence of two limit cycles bifurcating from the separatrix cycle in the quadratic
system, and Liu [43] three limit cycles bifurcating from the separatrix cycle in the
quadratic system (see also [78], [83] and [84]).
In Theorem 10, c1 , c2 , · · · are called separatrix quantities. Notice that divI(0) = 0
for Hamiltonian systems, we have
c1
c2
c3
=
(divI(ε))dxdy,
h≥0
= (divI(ε))|x=0,y=0 = σ0 (I(ε)),
=
(divI(ε))dt = σ1 (ε).
It is easy to see that c2 , c3 are just the criterion for determining the stability of the
separatrix cycle of system I(ε). The reason for this phenomenon is not clear. We
remark further that [30] is not only concerned with Hamiltonian systems, and the
methods presented in it are useful for further study.
Some of the separatrix quantities are related to the saddle point itself, some to
the separatrix cycle. Many people studied the so-called saddle quantities in [16], [18],
[19], [20], [21], [22] and [43] (see also [81]). Kokubu [43] pointed out that for some
special systems, the saddle quantities can determine the number of the bifurcation.
The authors of [63] showed that sometimes the saddle quantities can determine the
stability of the separatrix cycle. Luo and Zhu [22] stated that generally the saddle
quantities cannot be used to determine the stability of the separatrix cycle. Han [66]
also did some studies. Finally, Hu and Feng [72] and Han and Zhu [73] proved that
the saddle quantities whose order are less than two can determine the stability of the
separatrix cycle under suitable conditions. But for the higher order cases, there are
no better result. So it is interesting to study the role of the saddle quantities in the
stability and bifurcation of a separatrix cycle. Are the saddle quantities and focus
quantities the antithetic variables? Are there relations between the saddle quantities
and the separatrix quantities? All these are worthy of studying.
As for problem (4), from the above discussions, we see that it is difficult in general
to give the conditions for every possible bifurcation. But for some special cases, it is
possible. We have known that there are some limit cycles bifurcating from separatrix
cycle for a long time, but, until 1985, Feng and Qian [12] gave the analytic condition for
bifurcation, and Feng [17] gave the criterion for determining the limit cycle bifurcating
from heteroclinic orbit. For Hamiltonian systems, Li [37] weakened the condition in
[17], and gave the condition for the bifurcation of homoclinic cycle and heteroclinic
cycle. Feng and Xiao [38] gave the condition under which the homoclinic cycle and
heteroclinic cycle bifurcated from a heteroclinic cycle which extends the result in [37].
Recently, the attention of many researchers have turned to the homoclinic cycles or
heteroclinic cycles in Rn and some new problems appear. For example, the homoclinic
cycle in higher dimension system may have some bifurcations which are impossible
for planar system. The most remarkable case is the so-called homoclinic doubling
bifurcation, 2-impulse limit cycle and n-impulse homoclinic cycle and limit cycle (see
Figure 7).
B. Y. Feng and R. Hu
29
Figure 7
Deng obtained some conditions for homoclinic doubling bifurcation in [90], he gave
a homoclinic doubling bifurcation Theorem there. From the discussions mentioned
above, we should not only trace the information of the separatrix cycle itself as we do
in the planar system, but also study the invariant manifold near the separatrix cycle.
For instance, in order to determine whether there are homoclinic doubling bifurcations,
we must know whether the homoclinic cycle in the unperturbed system is twisted in the
sense that the two-dimensional surface formed by the stable manifold in a neighborhood
of the cycle is a Möbius band, but it is impossible to get the twisted property from the
homoclinic cycle itself. The required information must be obtained from the invariant
manifold (global not local) near the homoclinic cycle.
Recently, Sanstede [88] did some useful research works in this area, and he studied
the so-called hyperbolic center manifold, i.e. the invariant manifolds near the space
homoclinic cycle. [82], [86] and [89] are additional references in this area.
As for the invariant manifolds near the space homoclinic cycle, there remains a lot
of important studies to be done. For example, is it possible to get the analytic criterion
for twisted and non-twisted? In addition, it is clear that the space homoclinic cycle
contains more detailed information than the planar system does. For example, the three
kinds of bifurcations mentioned above (homoclinic doubling bifurcation, 2-impulse limit
cycle, and two limit cycles which are very closer) all tends to the same homoclinic cycle
as the perturbation parameter ε tends to zero. This fact means that under certain
conditions, there are at least 3 bifurcation directions for the space homoclinic cycle,
and there are at least 3 kinds of corresponding analytic criterion. What are these
criterion? What are the differences between them? These are interesting and unsettled
problems.
Figure 8
30
Homoclinic and Heteroclinic Orbits
Consider the system

 
ẋ
α −β
 ẏ  =  β α
ż
0 0




0
x
x
0  y  + f  y ,
λ
z
z
(12)
ω = α + iβ.
The following result is in [74] and [75].
THEOREM 11. If |α| < λ, β = 0 and when = 0, there exists a homoclinic orbit
γ (passing through O(0, 0, 0)), then there exists a perturbed system of (12) in which
there is a homoclinic orbit γ near γ, and the return map of γ includes horseshoes
which are countable (see Figure 9).
Figure 9
The homoclinic cycle mentioned above is called the Silnikov homoclinic cycle. The
eigenvalues of the system are λu > 0 and a pair of complex conjugate values λs and λ¯s
with Imλs = 0 satisfying λu > |Reλs |. In 1965, the authors of [74] and [75] showed the
following.
THEOREM 12. Assume that D is an appropriately defined domain of return map
π. For any positive integer m ≥ 2, there exists a Π-invariant subset Vm ⊂ D such that
Π|? Vm is topologically conjugate to the shift dynamics
+∞
σ : Sm = {1, · · · , m}Z → Sm ; (si )+∞
i=−∞ → (si+1 )i=−∞ .
In other words, it admits a homeomorphism h : Sm → Vm satisfying
Π|Vm ◦ h = h ◦ σ.
In [69], a proof for the above theorem is given. In 1990, Deng [76] gave a stronger
result. Deng [33] did further studies on the Silnikov problem. Sun [42] generalized the
result of [12] for the three dimensional case, and studied the bifurcation of the three
dimensional homoclinic orbit. Other results can be found in [95]. The results of [67]
showed that although it is difficult to study the higher dimensional cases, many results
are extensions of the two dimensional case. So many methods for the two dimensional
case are still useful for studying the higher dimensional cases.
Acknowledgment. The author is grateful to the reviewers, who make many helpful
suggestions, correct some of our mistakes and indicate several useful references [92]—
[107]. The author is also grateful to Professor Sui Sun Cheng. It is impossible to
complete this paper without his editorial help.
B. Y. Feng and R. Hu
31
References
[1] H. Dulac, On limit cycles, Bull, Soc. Math. France, 51(1923), 45-188.
[2] A. A. Andronov, E. A. Leontovich, I. I. Gordon, and A. G. Maier, Qualitative
Theory of Second-Order Dynamical System, Nauka, Moscow, 1966.
[3] A. A. Andronov, E. A. Leontovich, I. I. Gordon, and A. G. Maier, Theory of
Bifurcations of Dynamic Systems on a Plane, Wiley, New York, 1973.
[4] V. I. Arnold, Geometrical Methods in the Theory of Ordinary Differential Equations, Springer-Verlag, New York, 1982.
[5] S. N. Chow and J. K. Hale, Methods of Bifurcation Theory, Berlin, Springer
Verlag, 1982.
[6] J. Guckenheimer, P. Holmes, Nonlinear Oscillations, Dynamical Systems and
Bifurcation of Vector Fields, Springer-Verlag, New York, 1983.
[7] Y. Q. Ye, Theory of Limit Cycle(2nd edition), Shanghai Science and Technology
Press, Shanghai, China, 1984 (in Chinese).
[8] X. Wang, The global analysis of ẍ + (α1 + 3α3 x2 )ẋ − β1 x + β3 x3 = 0, Chinese
Mathematica Applicata, 2(1990), 59-64 (in Chinese).
[9] J. H. Sun, The uniqueness for the limit cycles of the quadratic system (III)δ=−m ,
Chinese Kexue Tongbao, 12(1990), 887-890 (in Chinese).
[10] J. W. Reyn, Generating of limit cycles from separatrix polygons in the phase
plane, Lecture Notes in Math., 810, 264-289.
[11] B. K. Melnikov, On the stability of the center for periodic perturbation of time,
Trans. Moscow. Math. soc., 12(1963), 1-57.
[12] B. Y. Feng and M. Qian, The stability of a saddle point separatrix loop and a
criterion for its bifurcation limit cycles, Acta Math. Sinica, 28(1985), 53-70 (in
Chinese).
[13] B. Y. Feng, Condition for generating limit cycles by bifurcation of the loop of
a saddle-point separatrix, Acta Math. Sinica, New series, 3(1987), 373-382 (in
Chinese).
[14] L. A. Cerkas, On the stability of a singular cycle, differential’nye uravnenija,
4(1968), 1012-1017.
[15] Z. E. Ma and E. N. Wang, The stability of a loop formed the separatrix of a
saddle point and the condition to produce a limit cycle, Chin. Ann. of Math.,
4A(1983), 105-110 (in Chinese).
[16] P. Joyal and C. Rousscau, Saddle quantities and applications, J.D.E., 18(1989),
374-399.
32
Homoclinic and Heteroclinic Orbits
[17] B. Y. Feng, On the stability of a heteroclinic cycle for the critical case, Acta
Math. Sinica, 33(1990), 113-134 (in Chinese).
[18] C. B. Qin and Y. X. Qin, Calculation of a class of saddle quantities, Chinese
Kexue Tongbao, 32(1987), 1841-1843 (in Chinese).
[19] Y. R. Liu and J. B. Li, The singularity values of complex autonomous differential
systems at singular points, Science in China, 3A(1989), 245-255 (in Chinese).
[20] S. L. Cai and G. Y. Guo, Saddle values and limit cycles generated by separatrix of
quadratic systems. Proceedings of Asian Mathematical Conference, 1990 (Hong
Kong), 25-31, World Sci. Publishing, River Edge, NJ, 1992.
[21] D. M. Zhu, Saddle-value and integrability condition of quadratic differential systems, Chin. Ann. of Math., 8B(1987), 466-478 (in Chinese).
[22] D. J. Luo and D. M. Zhu, The stability of homoclinic loop and the uniqueness
for generating limit cycles, Chin. Ann. of Math., 11A(1990), 95-103 (in Chinese).
[23] D. J. Luo, M. A. Han, and D. M. Zhu, The uniqueness of limit cycles bifurcating from a singular closed orbit (I), Acta Math. Sinica, 35(1992), 407-417 (in
Chinese).
[24] D. J. Luo, M. A. Han, and D. M. Zhu, The uniqueness of limit cycles bifurcating
from a singular closed orbit (II), Acta Math. Sinica, 35(1992), 4: 541-548 (in
Chinese).
[25] D. J. Luo, M. A. Han, and D. M. Zhu, The uniqueness of limit cycles bifurcating
from a singular closed orbit (III), Acta Math. Sinica, 35(1992), 5: 673-684 (in
Chinese).
[26] L. A. Cerkas, The stability of a special cycle for the critical case, Differential
Equations, 3(1967), 1060-1069 (in Russian).
[27] K. R. Zhou, Revision and supplement to a result of L. A. Cherkas, Sichuan Daxue
Xuebao, 2(1980), 23-27 (in Chinese).
[28] C. A. Holmes, Some quadratic systems with a separatrix cycles surrounding a
limit cycle, J. London Math. Soc. (2). 37(1988), 545-551.
[29] B. Aulbach and D. Flockerzi, The past in short hypercycles, J. Math. Biol,
27(1989), 223-231.
[30] H. K. Leontovich, On the generation of limit cycle from a separatrix, Dock.
Academic Science, CCCP, LXXVII, 4, 641-645.
[31] N. K. Gafrilov and L. P. Silnikov, On the three-dimensional dynamical systems
close to systems with a structurally unstable homoclinic curve I, Math. USSR
Sb., 88(4)(1972), 467-485.
B. Y. Feng and R. Hu
33
[32] N. K. Gafrilov and L. P. Silnikov, On the three-dimensional dynamical systems
close to systems with a structurally unstable homoclinic curve II, Math. USSR
Sb., 90(1)(1973), 139-156.
[33] B. Deng, The Silnikov problem, exponential expansion, strong λ-lemma, C 1 linearization and homoclinic bifurcation, J. D. E., 79(1989), 189-231.
[34] C. Z. Li and C. Rousscau, Codimension 2 symmetric homoclinic bifurcation, Can.
J. Math., 42(2)(1990), 191-212.
[35] D. Stoffer, On the approach of Holmes and Sanders to the Melnikov procedure
in the method of averaging. Z. Angew. Math. Phys. 34 (1983), no. 6, 948—952.
[36] H. Kokubu, Homoclinic and heteroclinic bifurcations in vector fields, Japan J. of
Appl. Math., 5(1988), 455-501.
[37] B. Y. Li, Bifurcation phenomenon of the heteroclinic cycle in Hamiltonian system
with autonomous perturbation, Acta Math. Appl. Sinica, 2(1994), 239-247 (in
Chinese).
[38] B. Y. Feng and D. M. Xiao, Generation of heteroclinic cycles from a heteroclinic
cycle and homoclinic cycle by bifurcation, Acta Math. Sinica, 35(1992), 815-830
(in Chinese).
[39] B. Y. Feng, The Melnikov criterion for an infinite separatrix loop and the distribution of limit cycles in a quadratic system, Chinese Acta Math. Appl. Sinica,
16(1993), 482-492 (in Chinese).
[40] X. F. Yuan, Second-order Melnikov functions and their applications, Acta Math.
Sinica, 37(1994), 135-144 (in Chinese).
[41] B. Y. Feng, The stability of a heteroclinic cycle for the critical case, Scientia
Sinica, 34A(1991), 920-934 (in Chinese).
[42] J. H. Sun, Bifurcation and chaotic dynamics of homoclinic systems in R3 , Acta
Math. Sinica (new series) 11(1995), 128-136.
[43] Y. R. Liu, The values of singular point of En and some kinds of problems of
bifurcation, Sci. China., Ser. A, 36(1993), 550-560 (in Chinese).
[44] H. Kokubu, Y. Nighiura and H. Oka, Heteroclinic and homoclinic bifurcation in
bistable reaction diffusion systems, J.D.E., 86(1990), 260-341.
[45] S. Van Strien, Smooth linearization of hyperbolic fixed point without resonance
condition., J.D.E., 85(1990), 66-90.
[46] P. Joyal, The cusp of order n, J.D.E., 99(1990), 1-14.
[47] G. Wolansky, limit theory for a dynamical system in the presence of resonances
and homoclinic orbits, J.D.E., 83(1990), 300-335.
34
Homoclinic and Heteroclinic Orbits
[48] J. Sotomayor, Curvas Definidas for Equagoes Diferencials No Plano, Instituo De
Matematica Pura e Aplicada, 1981.
[49] The Singular Points and Limit Cycles of the Differential Equations on a Real
Plane and Complex Plane, as the appendix of Curvas Definidas for Equagoes
Diderencials No Plano (in Chinese ).
[50] J. H. Sun, Bifurcation to separatrix loops in supercritical cases, Chin. Ann. of
Math., 12A(1991), 636-643 (in Chinese).
[51] J. H. Sun, On the three dimensional homoclinic systems(I), preprint (in Chinese).
[52] J. H. Sun, On the three dimensional homoclinic systems(II), preprint (in Chinese).
[53] R. Roussarie, On the number of limit cycles which appear by perturbation of
separatrix loop of planar vector fields, Bol. Soc. Bras Mat., 17(1986), 67-101.
[54] J. Guckenheimer, R. Rand, D. Schlomiuk, Degenerate homoclinic cycles in perturbations of quadratic Hamiltonian systems, Nonlinearity, 2(1989), 405-418.
[55] F. Dumortier, R. Roussarie, J. Sotomayor, Generic 3-parameter families of vector
fields on the plane, unfolding a singularity with nilpotent linear part, The cusp
case of codimension 3, Ergodic Theory and Dynamical Systems, 7(1987), 375-413.
[56] Y. R. Zhou, The existence of limit cycles of nonlinear oscillating equations, Acta
Math. Appl. Sinica, 3(1980), 50-56 (in Chinese).
[57] S. X. Yu, On Filippov’s method in the proof of existence and uniqueness of limit
cycles, Acta Math. Sinica, 14(1964), 461-470 (in Chinese).
[58] P. G. Zhang, The stability of separatrix cycle S ( 2) with two singularities for a
quadratic systems, Chin. Ann. of Math., 11A(1990), 361-365 (in Chinese).
[59] B. Q. Shen, The problem on the existence of cubic curve limit cycle and cubic
curve separatrix cycle in quadratic system, Chin. Ann. of Math., 12A(1991),
382-389 (in Chinese).
[60] P. Hartman, Ordinary differential equations, Boston, Basel, Stutgart, Birkhauser,
1982.
[61] S. Sternberg, Lectures on Differential Geometry, Prentice-Hall, 1964.
[62] M. C. Irwin, Smooth Dynamical Systems, Academic Press, Inc., London, 1980.
[63] J. Palis, Jr, Welington de Melo, Geometric Theory of Dynamical Systems: An
Introduction, Springer-Verlag, New York, Heidelberg, Berlin, 1982.
[64] C. Rousseau, Example of a quadratic system with two cycles appearing in a
homoclinic loop bifurcation, J.D.E., 66(1987), 140-150.
B. Y. Feng and R. Hu
35
[65] G. R. Dai and Z. J. Leng, The separatrix cycles and integral straight lines of
E3 -systems, Journal of Sichuan University Natural Science Edition, 24(1987),
367-375 (in Chinese).
[66] M. A. Han, The number of limit cycles bifurcating from homoclinic and heteroclinic singular closed orbits, Science in China, A23(1993), 113-122 (in Chinese).
[67] D. M. Zhu, Homoclinic bifurcation with codimension 3, Chin. Ann. of Math.,
Ser. B, 15(1994), 205-216 (in Chinese).
[68] P. Joyal, Generalized Hopf bifurcation and its dual generalized homoclinic bifurcation, SIAM. J. Appl. Math, 48(1988), 481-496.
[69] T. Matsumoto, M. Komuro, H. Kokubu, R. Tekunaga, Bifurcation, Sights,
Sounds and Mathematics, Tokyo, Springer-Verlag, 1993.
[70] B. Y. Feng, Stability and bifurcations of a separatrix tending to infinite, Acta
Math. Sinica, 38(1995), 682-695 (in Chinese).
[71] B. Y. Feng, On the stability of a homoclinic cycle and a heteroclinic cycle in
space, Acta Math. Sinica, 39(1996), 649-658 (in Chinese).
[72] R. Hu and B. Y. Feng, The Criterion for determining the semi-stability of a
multiple limit cycle and the stability of a homoclinic cycle for the second critical
case, preprint.
[73] M. A. Han and D. M. Zhu, The Bifurcation Theory of Differential Equation,
China Coal Industry Press, Beijing, China, 1994 (in Chinese).
[74] L. P. Silnikov, A case of the existence of a countable set of periodic motions, Sov.
Math. Dokl., 6(1965), 163-166.
[75] L. P. Silnikov, The existence of a denumerable set of periodic motions in a fourdimensional space in an extended neighborhood of a saddle-focus, Sov. Math.
Dokl., 8(1967), 54-58.
[76] B. Deng, On Silnikov’s homoclinic-saddle-focus theorem. J. D. E., 102(1993),
305-329.
[77] J. B. Li and B. Y. Feng, Stability, Bifurcations and Chaos, Yunnan Science and
Technology Press, Kunming, China, 1995 (in Chinese).
[78] Z. F. Zhang, C. Z. Li, Z. M. Zheng, and W. G. Li, The Basic Bifurcation Theories
of the Vector Field, China Higher Education Press, Beijing, China, 1997 (in
Chinese).
[79] J. Y. Zhang and B. Y. Feng, The Geometrical Theory and Bifurcation Problem
of Ordinary Differential Equation(the second revised edition), Peking University
Press, Beijing, China, 2000 (in Chinese).
36
Homoclinic and Heteroclinic Orbits
[80] B. Y. Li, Z. F. Zhang and C. Z. Li, Bifurcation phenomenon of a class of degenerate quadratic Hamiltonian systems with quadratic perturbations, Journal of
Nonlinear Dynamics in Science and Technology, 2(1995), 345-355 (in Chinese).
[81] D. M. Zhu, Homoclinic bifurcation value, preprint (in Chinese).
[82] D. M. Zhu, Problems in homoclinic bifurcation with higher dimensions. Acta
Math. Sinica, 41(1998), 1289-1294 (in Chinese).
[83] M. A. Han, Cyclicity of planar homoclinic loops and quadratic integrable systems,
27A(1997), 891-902 (in Chinese).
[84] X. P. Li, A kind of two-parameter cubic curve family defined by a quadratic
system, Chin. Ann. of Math., 9A(1998), 677-688 (in Chinese).
[85] W. G. Li, Normal Form Theorem and Application, Science Press, Beijing, China,
2000 (in Chinese).
[86] Q. P. Tian and M. A. Han, Non-twisted bifurcation of heteroclinic cycle, Science
in China, A30(2000), 193—202 (in Chinese).
[87] M. A. Han and S. S. Gu, The Theories and Methods of Nonlinear System, Beijing,
China, 2001 (in Chinese).
[88] B. Sandstede, Center Manifolds for Homoclinic Solutions, preprint.
[89] A. Stephan, V. Gils, M. Krupa and V. Tshistiakov, Homoclinic twist bifurcation
in a system of two coupled oscillator, J. Dynam. Diff. Eq., 12(2000), 733-806.
[90] B. Deng, Homoclinic twisting bifurcations and cusp horseshoe maps, J. Dynam.
Diff. Eq., 3(1993), 417-467.
[91] M. A. Han and L. Y. Li, The uniqueness of limit cycles bifurcating from a homoclinic cycle and heteroclinic cycle, Chin. Ann. of Math., 16A(1995), 645-651 (in
Chinese).
[92] S. Smale, Diffeomorphisms with many periodic points, in Differential and Combinatorial Topolopy, Princeton University Press, Princeton.
[93] S. Smale, Differentiable dynamical system, Bull. Amer. Math. Soc., 73(1967),
747-871.
[94] W. G. Li, Normal Form Theory and It’s Application, Science Press, China, 2000
(in Chinese).
[95] I. Yu and W. G. Li, Nonlocal Bifurcation, Mathematical Surveys and Monographs, Vol. 66, AMS, 1999.
[96] F. Battelli and C. Lazzari, Exponential dichotomies, heteroclinic orbits and Melnikov functions, J. Diff. Eq., 86(1990), 342—366.
B. Y. Feng and R. Hu
37
[97] S. N. Chow, J. K. Hale and J. Mallet-Paret, An example of bifurcation to homoclinic orbits, J. Diff. Eq., 37(1980), 351—373.
[98] M. Fečkan, Higher dimensional Melnikov mappings, Math. Slovaca, 49(1999),
75—83.
[99] J. Gruendler, Homoclinic solutions for autonomous ordinary differential equations
with nonautonomous perturbations, J. Diff. Eq., 122(1995), 1—26.
[100] J. Knobloch, Bifurcation of degenerate homoclinics in reversible and conservative
systems, J. Dynam. Diff. Eq., 9(1997), 427—444.
[101] G. Kovačič, Singular perturbation theory for homoclinic orbits in a class of nearintegrable dissipative systems, SIAM J. Math. Anal., 26(1995), 1611—1643.
[102] X. B. Lin, Using Melnikov’s method to solve Silnikov’s problems, Proc. Royal
Soc. Edinburgh 116A(1990), 295—325.
[103] D. W. McLaughlin, E. A. Overman II, S. Wiggins and C. Xiong, Homoclinic orbits in four-dimensional model of a perturbed NLS equation: a geometric singular
perturbed study, Dynamics Reported 190—287, Dynam. Report. Expositions Dynam. Systems (N.S.), 5, Springer, Berlin, 1996.
[104] K. R. Meyer and G. R. Sell, Melnikov transforms, Bernoulli bundles, and almost
periodic perturbations, Trans. Amer. Math., 314(1989), 63—105.
[105] K. J. Palmer, Exponential dichotomies and transversal homoclinic points, J. Diff.
Eq., 55(1984), 225—256.
[106] A. Vanderbauwhede, Bifurcation of degenerate homoclinics, Results Math., 21(12)(1992), 211—223.
[107] S. Wiggins, Global Bifurcations and Chaos, Analytical Methods, Applied Mathematical Sciences, Vol. 73, Springer-Verlag, New York, 1988.
[108] N. J. Balmforth, Solitary waves and homoclinic orbits, Annual review of fluid
mechanics, Vol. 27, 335—373, Annual Reviews, Palo Alto, CA, 1995.
[109] R. M. May and W. Leonard, Nonlinear aspect of competition between three
species, SIAM J. Appl. Math., 29(1975), 243—252.
[110] J. Hofbauer and K. Sigmund, On the stabilizing effect of predator and competitors
on ecological communities, J. Math. Biol., 27(1989), 537-548.
[111] B. Y. Feng, The heteroclinic cycle in the model of competition between n species
and its stability, Acta Math. Appl. Sinica, 14(1998), 404—413.
Download